Bidimensional Structure Data Mining Ordering Prediction Approximation
Issue Date:
2004
Publisher:
Institute of Information Theories and Applications FOI ITHEA
Abstract:
The method (algorithm BIDIMS) of multivariate objects display to bidimensional structure in which the
sum of differences of objects properties and their nearest neighbors is minimal is being described. The basic
regularities on the set of objects at this ordering become evident. Besides, such structures (tables) have high
inductive opportunities: many latent properties of objects may be predicted on their coordinates in this table.
Opportunities of a method are illustrated on an example of bidimentional ordering of chemical elements. The table
received in result practically coincides with the periodic Mendeleev table.